Cognitive Overload: Jailbreaking Large Language Models with Overloaded Logical Thinking

arXiv:2311.09827v2 Announce Type: replace
Abstract: While large language models (LLMs) have demonstrated increasing power, they have also given rise to a wide range of harmful behaviors. As representatives, jailbreak attacks can provoke harmful or unethical responses from LLMs, even after safety alignment. In this paper, we investigate a novel category of jailbreak attacks specifically designed to target the cognitive structure and processes of LLMs. Specifically, we analyze the safety vulnerability of LLMs in the face of (1) multilingual cognitive overload, …

abstract alignment arxiv attacks cognitive cs.cl jailbreak jailbreaking language language models large language large language models llms novel overload paper power responses safety thinking type